Friday, April 27, 2012

A few answers to Bright Knight

BK said:
It is not at all clear that it is easier to improve by a fixed percentage on harder problems than it is on easier ones. Back in the days of the Bain Experiment, I found that reducing my solution times by a fixed percentage for my first pass through the first batch of problems gave a good fit to my solution times on the first passes through the later batches of problems. That appears to be roughly right for Woolum too.

I advocate to train every specific task that you have to perform in a specialized way. Optimized for that specific task. Preventing diminishing returns.

The first thing to do is to determine which tasks we are talking about. Every tasks has its own specialized sets of patterns associated with it, and if you don't find out which patterns to learn, chances are high that you don't learn these patterns at all.

First task: 20 tactical elements.
The first task we have identified is that we have 20 - 30 elementary tactical themes, depending on the source. Let's round it off to 20 usefull themes. In my opinion these are trained best according to my speedtraining method:
Make per tactical element a separate problemset with rating <1200. Solve these problems as fast as possible, untill you reach a speed of <4 seconds per problem. 1000x per tactical element is probably enough to recognize it everywhere and fast. Let's say a problem takes 20 seconds at average. This leads to a total time of 20 elements x 1000 times x 20 seconds =  111 hours to master this task.

Doing the same with combinations I consider to be useless, since there are way too much combinations.

Doing problems higher rated than 1200 is useless for this task. Here the law of diminishing results sets in soon.

Second task: guiding your mind for duplo attacks.
I identified duplo attacks as one of the three ways to gain wood. Traps and promotion being the other two. This means that you have to scan a position for all its duplo attacks. There are 5 different duplo attacks (double attack, discovered attack, pin, skewer, rontgen attack). The recognition itself of these 5 patterns is already taken care of when training for the first task. You only have to make it a habit to scan for all of these duplo attacks in stead of stopping short once you get distracted by one of them which looks promising.
I use high rated problems for this for a starter. Once I have identified all duplo attacks I store both the position and the duplo attacks as a flashcard in Anki and train them further with the aid of Anki, without going back to CT anymore.

Third task: making one of the duplo attacks work.
At average there are about 6 duplo attacks in a >2000 rated problem. Only one of them is winning. At this stage is is my task to find out which one and how I can make it work. For this I need all tactical elements that I have learned already in task one and that can act as some prelimanary move. Besides that, some conscious thinking is probably needed here in order to make it work.

Fourth task: identifying the focal squares.
The fourth task is needed to assist task three.
Once I know the target, I must find the way back to the attacker, finding the focal points along the road. These focal points give me clues to the preliminary moves I ave to make before the combination will work. I must identify the defenders of the focal points and find a way to annihilate them. I must identify where my own pieces get in the way and clear them, etc..
I use high rated problems for a starter. Once I have identified the focal points I make a flashcard with the position and the focal points. From there I train it with Anki.

Fifth task: learning the mating patterns.
This has great similarity with task one.
If I'm not mistaken there are 29 different mating patterns that must be learned. Speedtraining with <1200 rated problems will do the job.

Sixth task: learning the patterns that are associated with promotion.
This has great similarity with task one.

Speedtraining with <1200 rated problems will do the job.

Seventh task: Chasing the king.
I just discovered this as a problem. It has its own associated patterns. First I must formulate a search algorithm. In order to find one I use high rated problems at CT which are mates. Usually> 4 movers.
Only once I have found a suitable algorithm, I can start to think about a method how to train them.
To give you an idea of the complexity of the matter: a search algorithm must guide you through this testposition in order to be usefull.

Eight task: positional play.
Under construction. See previous posts.

In summary:
Speed training with tagged low rated problems is suitable for most elementary patterns. High rated problems are used in two cases:
  • As a starter to extract patterns that you will only find in complex problems. The actual learning is done with flashcards and Anki.
  • During investigation. When you are investigating new tasks and defining new search algorithms. This is only temporarily.

It is true that you have an average of about 3 minutes per move in a standard time limit game, but how much time can you afford to spend on checking that your chosen move is proof against a tactical shot? I would have thought that 1 minute would be optimistic. Hoping for a sixty fold improvement in your solution time is too optimistic in my opinion.

That's all way too ridgid and too theoretical. We will find this out in an empyrical way.

What failure rate is acceptable when checking for tactical shots? 10%, 1% or less… There are typically about 40 moves in a game, and many or most of them could fall foul of tactical shots. If you accept a 10% failure rate, you are going to “blunder” once or more in most games. Even if you get your failure rate down to 1%, you will regularly lose to tactical shots.

Again way too theoretical. Not all positions call for an ability to solve tactics at 2300 level. But sometimes tactics just must work or you loose. If I look at the Polgar middlegame brick, I could flaw about 25% of the tactics that were presented there with the aid of Rybka. This means that both the grandmasters who played the game, the grandmasters that annotated the game in the Informator and Susan Polgar who checked the problems for the book overlooked the flaws. When both you and your opponent miss the same tactic, the tactic is irrelevant for the game.


  1. The 80% correlation between blitz and standard time limit performance also suggests that both are equally difficult to improve beyond a certain point.

    I meant the last two paragraphs to follow sequentially, rather than to be taken in isolation. Yes, difficult tactics that neither player spots are practically irrelevant. The point I was making is that what is most important is not to fall to simple tactical shots, and to capitalise on them when they are available to you. High success rate in 1 minute implies easy problems.

    Diminishing returns are inevitable, but there is more to chess than tactical problems!

    How did the top players get that good? In the past, there were no computers, so the answer was simply study books and play chess. There were very few problem books available in days gone by. How did Morphy train?

  2. @BK,
    The point I was making is that what is most important is not to fall to simple tactical shots, and to capitalise on them when they are available to you. High success rate in 1 minute implies easy problems.

    To become a master, you need to be able to solve masterlevel problems with a succesrate of 50%.

    The fact that you are able to solve low rated problems with a succesrate of >99% doesn't change that.

    That what is typical for high rated problems can't be learned from low rated problems since the patterns aren't there.

    How did Morphy train?

    I don't know. But even if I did, I question that it is relevant. Since we are adults and he was young when he learned the game.

  3. I am not saying that you should train only on the very easiest stuff, just that mastery of the easy stuff has more value than an incomplete understanding (or misunderstanding) of the complicated stuff. Learn to walk before you try to run! To become a master I have to learn lots of things. That takes time. Pick the low lying fruit first! I believe that the learning process is the same young or old. Morphy became a very strong tactical player with a few books written by weaker players and practice against weaker players. I am not saying that is optimal, but it worked.

  4. I consider the readers of this blog to be ready with harvesting low hanging fruit. Otherwise I suggest to harvest first the low hanging fruit:)

  5. Here is a Dan Heisman article that puts the case for easy problems well:

    Coakley emphasises the importance of training for accuracy rather than speed, but the two can be traded off to some extent. One way of getting more accurate is to remember things rather than having to work them out, which benefits both.

    If it is obvious that tactics are in the air, you may be able to spend several minutes looking for them in a game. With “bolt from the blue “ tactics, you will be doing well if you can afford to look for a minute, particularly if you are on the receiving end. I believe that it is fine to routinely spend a few minutes on a problem. Solving a problem is more instructive than being told the solution. Much more than that? Occasionally, perhaps.

  6. Both Morphy and Anderssen, huge tacticians back in their day, created tactical puzzles, wrote columns and went over games.

    To create your own tactical problem from scratch is interesting because you have to have deep knowledge of the themes that you spell out in this post as well as "know" what makes a strong combination to challenge varying degrees of readers.

  7. Given lately that my life takes on certain tangents that takes me away from the chess board, I have been deliberate in training in ways to minimize skill loss and still find a way to improve albeit sporadic intervals and levels of chess commitment. So, here I am coming back from an intense 5 month hiatus with absolutely no chess involvement. I am looking at my re-entry as an experiment in retention of skills from a cognitive perspective as much as possible as well as identify which are I lost the most and come up with a plan to “rust proof” that area. In reality it’s been about 8 months since I’ve really indulged heavily in the chess improvement circles and resurfaced briefly during Holiday breaks.

    I am looking at the following areas of where my skills started to disappear: 1) tactics 2)Positional 3) Openings, and 4) endgames. I will save more of the details for a post on my blog.
    Your post made me ponder the different layers of item 1, tactics, and where I may have lost the most.

    Because of the past MDLM training and the unfocused 7 circles of hell , the best that came about from all that thrashing was the rudimentary tactics are still etched in my brain. After not looking at a chess board for months, anything beyond what you call “First Task” was challenging. I recognized from the start why this was the case. When I stormed the seven circles, I never took the time to create that mental dialog for the basic elements. “That one is a removal of the guard” “This one is a geometrical motif with a knight” etc. So as I “warm up” this time with the first task, I am making sure I do the deliberate mind mapping necessary to transfer the skill from a memory function to an actual skill in long term memory with a memory marker that I can recall through us of my own narrative.

    I did this with positional motifs as a means to understanding my opening repertoire better and forcing myself to let the position dictate candidate moves than rely solely on rote memorization of move orders. It helped back in December when after not playing for several months, I was still able to play a decent game and not drop (much) in skill level.

    So if I can apply this technique of creating memory maps through the right cognitive internal dialog in the four areas, then I stand a better chance of developing “chess skills” that are meant to last.

  8. @BP,
    I look forward to your post.

    With hindsight, the path is pretty obvious. Strange that it takes so long to see it clear.

  9. "How did Morphy train?"

    He read over annotated games, sitting in his gazebo, with no board in sight.

    I saw a short clip of Magnus Carlson at London. Hikaru moves, Magnus writes down the move and stares past his scoresheet, then makes a move on the board. Clearly he is playing the game blindfold and barely glanced at the board just to make the move.

  10. Who really knows how Morphy trained. One thing is for sure Morphy probably would have killed to have the resources we have today to learn from. I'm a bit skeptical that all this defining of tactical problems will really do all that much for you. Just notice little things like open king position, pinned piece, weak square, that sort of thing and you'll spot tactics quickly. Whatever works for you though, but I'd much rather look at more Botvinnik games where there's plenty of tactics and awesome positional instruction...

  11. Who really knows how Morphy trained. One thing is certain; he would have killed to have the resources we have today.